microsoft/onnxruntime

Pull Requests Commits

Fix packed zero-point row indexing

Copilot committed 1 day ago

Verified 3ed68391

Fix GatherBlockQuantized CUDA block indexing

Copilot committed 2 days ago

Verified d1875aea

Copilot committed 2 days ago

Verified c6a507f6

Restrict XQA CUDA compilation to SM80 and newer (#32547)

Copilot committed 2 days ago

Verified 0db6c0e8

Improve CPU provider build time with Eigen PCH (#32616)

titaiwangms committed 2 days ago

Verified f9b87e94

Add BF16 routing support to CUDA MoE (#32659)

kunal-vaishnavi committed 2 days ago

Verified e266b17e

Add complete GQA workspace route recipes (#32454)

titaiwangms committed 2 days ago

Verified 8808c862

Optimize CPU uint8 GatherBlockQuantized dequantization (#32621)

Copilot committed 2 days ago

Verified 180041af

Validate NCHWc convolution layouts (#32557)

apsonawane committed 2 days ago

Verified 460dee98

Fix --enable_arm_neon_nchwc being a silent no-op on Windows ARM64 (#32094)

Copilot committed 2 days ago

Verified f38538cd

[CUDA] Add opt-in SM90 DeepGEMM for MatMulBlockQuantizedFp8Weight (#32599)

tianleiwu committed 2 days ago

Verified 29874882

Minor follow-up to GQA value tensor layout changes (PR 32139) (#32519)

javier-intel committed 2 days ago

Verified 30e26321

[WebGPU] Move several elements per thread in Concat, Slice, Transpose and Pool (#32497)

linzuojian committed 2 days ago

Verified 2a720f92

[CUDA] Make GatherND capture-safe (#32120)

tianleiwu committed 3 days ago

Verified 71b33460

Specify GroupQueryAttention windowed KV cache layout contract (#32419)

tianleiwu committed 3 days ago

Verified cd9da083

[CUDA] Add cuDNN paged SDPA decode tier for PagedAttention (#32493)

hariharans29 committed 3 days ago

Verified 4725a200

Add FP8/FP4 block-quantized support to GatherBlockQuantized contrib op (#32480)

Copilot committed 3 days ago

Verified aad07ef2

Suppress preload errors for optional CUDA libraries (#32112)

Copilot committed 3 days ago

Verified 84d758f8

Skip compiling Flash Attention .cu sources when onnxruntime_USE_FLASH_ATTENTION=OFF (#32580)

Copilot committed 3 days ago

Verified e8787e28

Add Qwen4-Exp EngramGate and NGramHashMapping support (#32285)

Copilot committed 3 days ago

Verified d71d2789

[MLAS] Fix Inf/NaN comparison and add fp16 extreme-value coverage for ARM64 HGEMM NEON (#31761)

kjg0724 committed 3 days ago

Verified 6f6e8815

Add GQA XQA and Flash workspace recipes (#32453)

titaiwangms committed 3 days ago

Verified 8ed3bea2

Start transitioning pools to unified Windows pool (#32612)

eserscor committed 3 days ago

Verified 1474dfc6

[WebGPU] Fix OOB store in 8x16x16 subgroup-matrix MatMulNBits (#32582)

jchen10 committed 3 days ago

Verified 3578d771

Add contributor sorting option to compile_contributors.py (#32549)

tianleiwu committed 3 days ago

Verified c96cf8de

Enable WebGPU GQA local-window flash attention (#32462)

feich-ms committed 3 days ago

Verified fed7f01a

[WebGPU] Vectorize NHWC depthwise convolution across channels (#32498)

linzuojian committed 3 days ago

Verified 73534a26

[WebGPU EP] implements VarlenCausalConvWithState operator (#32477)

prathikr committed 4 days ago

Verified d6750b39

[CUDA] Fix SM120 QMoE MoE GEMM TMA build failure with GCC (#32584)

bheu committed 4 days ago

Verified 5f44b5b5

[Telemetry] Reduce 1DS event volume (#32425)

bmehta001 committed 4 days ago

Verified 0a82f2ca

Older